AppLovin
An automated ad broker for phone games: in milliseconds it decides which advertisement each player sees, and keeps a cut of every one it places.
Our viewAttractive Low conviction
The scores
The profile · 1-4
Growth3/4 ↓
Three years of roughly 70% growth, none of it bought - but management names one driver, and the quarter it failed sales barely moved.
Business quality3/4 ↓
A data-scale advantage an independent attribution firm ranks first in tier-one markets, but the named rival's model has accelerated four quarters running.
Financial position4/4 →
Net debt of 0.08 times EBITDA, nothing due before 2029, an undrawn billion-dollar facility, and every plan funded from operations.
Execution record3/4 ↓
Five straight beats, then a quarter below its own guided EBITDA floor and unflagged - one slip, not yet a pattern.
Per share capture4/4 →
Share count fell from 338 to 335 million in six months; buybacks ran 8.9 times stock compensation and no convertible exists.
Potential2/4 →
The pond is large and nothing in it is signed; the consumer route needs creative tools, data density and awareness it does not yet have.
Risk profile3/4 →
Several disclosed concentrations - mobile gaming, three distribution platforms, 67% voting control, a live class action - and none of them binary.
Valuation4/4 →
At 24.5 times free cash flow and a 4.1% yield against a six-year median of 1.8%, the price pays for the current business alone.
Evidence quality4/4 →
Both filings, both releases, the call and the proxy, six outside filers, and two outside measurers for the moat, the threat and the prize.
The bet · 2-3 years
Asymmetric2.2×
If it works 2.2× · if it fails you keep about 100%.
The Verdict
We rate AppLovin attractive. It runs an automated ad broker for phone games and other apps, one of the most profitable businesses we cover, so the business is solid. In June its growth nearly stalled and the shares halved; the price now assumes that stall is permanent, so it is cheap, the cheapest of its rivals on spare cash. We think the stall was mostly timing. As a bet it is safe, with little upside: gaming ads alone cover today's price, so online shop ads come free. The view hinges on the September quarter landing inside the promised range. Conviction is low.
The long caseIf online shops follow game makers into AppLovin's auction, the same machine that matched games to players can sell their products to a billion daily users.
The evidence under this - 11 claims
C08facthigh confidence
Q2 2026 income from operations was $1,494,277 thousand on $1,923,686 thousand of revenue: an operating margin of 77.7%, against 76.1% in the same quarter of 2025.
C07facthigh confidence
AppLovin ran a business generating $5,480,717 thousand of FY2025 revenue with 898 employees and $147,902 thousand of long-lived assets, down from $196,443 thousand a year earlier.
C21facthigh confidence
Q2 2026 revenue of $1,923,686 thousand grew 52.8% year on year but only 4.4% sequentially, against 11.1% the quarter before and 18.0% the quarter before that.
C54facthigh confidence
The shares fell 50.0% from $733.60 on 22 December 2025 to $366.91 on 12 February 2026, 22.4% of it in the session after a fourth-quarter print that beat, on objections about Meta, Google and a short report; after the first-quarter print they closed at $613.70 on 1 June.
sourceFMP end-of-day price history for APP, 15 April 2021 to 22…primaryThe Motley Fool, 'Why AppLovin Stock Fell 24% in the First…secondaryAppLovin first quarter 2026 results release, Exhibit 99.1 to…primary
against this
- The coverage names three objections at once and a software sell-off beside them, so the February fall cannot be pinned on the one objection this thesis is about.
- A recovery says the market changed its mind after the next print; it does not say the objection was wrong, and the same shares then halved again from 1 June.
C36facthigh confidence
The shares trade at 25.2 times trailing earnings against a six-year median of 54.0, and at a 4.1% free cash flow yield against a six-year median of 1.8%; the company's own releases put free cash flow at $3.95 billion in 2025 and $863.3 million in the June quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primaryAppLovin fourth quarter and full year 2025 results release,…primary
- 2026-09-23editedMultiples refreshed with the pack on 23 September 2026, at a price 10.1% higher.
C22management claimmedium confidence
Management attributes the sequential slowdown to timing in its own research output: the pace of meaningful model improvement was lighter than normal and the next step-up landed just after quarter end.
against this
- It is the explanation most favourable to the company, offered by the company, for the first guidance-midpoint miss in its public history, and no filed document corroborates the timing of a model release.
- Bank of America had already flagged in July that new e-commerce sign-ups cooled in June, which is a demand-side account of the same quarter and is not what management describes.
- A complaint filed in September 2026 puts a competing account of the same quarter on the record: it alleges the constancy of the model improvement was overstated throughout a class period ending on the day of that print. It is an allegation and decides nothing, and it is the second named account of this quarter that is not management's.
- Unity's advertising segment, running on its rival model, grew 10.6% sequentially in the same quarter against AppLovin's 4.4% - a filed figure for a competitive account of the quarter, where the others are an analyst note and a pleading.
- The chief technology officer since 2020 stepped down on 1 July, the day the quarter closed; the handover was announced in April and the successor already ran engineering, so it explains nothing by itself, but management's account of the quarter does not mention it.
- 2026-09-21editedThe counter-evidence held one outside account of the quarter. A second class action now alleges the opposite of what management said about the model cadence, which belongs beside it.
- 2026-09-23editedTwo facts about the same quarter had not been read against the claim: a filed rival figure that grew faster, and a change of technical leadership on the day it closed.
C58inferencehigh confidence
Unity's Grow Solutions grew 6.4%, 10.9%, 23.5% and 35.4% year on year in the four quarters to June 2026, to $388,850 thousand, crediting Unity Vector; in the June quarter it grew 10.6% sequentially while AppLovin grew 4.4%.
sourceUnity Software Form 10-Q, quarter ended 30 June 2026primaryUnity Software Form 10-Q, quarter ended 30 September 2025primaryUnity Software Form 10-K, fiscal year ended 31 December 2025primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- AppLovin's revenue is 4.95 times Unity's whole segment and grew faster year on year in every one of the four quarters - 52.8% against 35.4% in the last.
- One quarter is one quarter: in the first quarter of 2026 Unity grew 4.0% sequentially against AppLovin's 11.1%, so the June reading may be Unity's own lumpiness as much as a shift.
C37facthigh confidence
At 24.5 times free cash flow AppLovin is the cheapest of the five companies its own annual report names, against a cohort median of 57.7 - Unity 35.0, Meta 45.8, Digital Turbine 69.6, Alphabet 79.1. Its own free cash flow is filed: $3.95 billion in 2025.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin fourth quarter and full year 2025 results release,…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primary
against this
- On sales it is the dearest of the same five at 16.2 times against a median of 8.75, so the answer inverts with the metric and the choice of metric is doing the work.
- A free cash flow multiple flatters a company whose free cash flow has just grown 58% and would look very different against the flat figure the bear case expects.
- 2026-09-23editedCohort refreshed with the pack on 23 September 2026: AppLovin still ranks cheapest, and the peer figures moved.
C39inferencemedium confidence
Three brokers cut the stock within a week of the print and the shares fell 19.7% the day after it, 6.0% five sessions later - the price move is a repricing of the growth rate, not of the quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- The July fall came first and on a different cause - an analyst note flagging cooled e-commerce sign-ups in June - so the drawdown is at least two events and only one of them is the print.
- A stock priced at 54 times earnings does not need a large miss to lose a fifth of its value; the magnitude is a function of where the multiple started.
C67estimatelow confidence
Taking AppLovin's trailing free cash flow at about $4.5 billion (24.5 times on a $110.4 billion market value) and removing roughly $0.75 billion of consumer sales at the company's own cash margin of about 66 cents per sales dollar leaves about $4.0 billion of free cash flow for gaming ads and MAX.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primary
against this
The consumer split is not disclosed (C65); if consumer sales are larger than assumed, the core is smaller and the option less free.
C65management claimlow confidence
Management said in May 2026 that the consumer business had passed a $1 billion run rate of spend a year earlier and is 'much bigger now', with April larger than any month of the December quarter; in February it declined to report e-commerce apart from gaming.
sourceAppLovin first quarter 2026 earnings call (FMP), the remarks…primaryAppLovin fourth quarter 2025 earnings call (FMP), the answer…primary
against this
A run rate of advertiser spend is not revenue, which is spend less what publishers are paid, and 'much bigger' is not a figure; nothing filed corroborates either.
The company
App developers find new players through advertisements shown inside other people's games. AppLovin runs the auction that places them.
An advertiser tells it what return it needs. Its software then decides which advertisement each player sees, and at what price. Without it, the advertiser buys by hand on Meta, Google or Unity and guesses the price itself. After paying for the computers, AppLovin keeps about 88 cents of every dollar it charges.
The evidence under this - 5 claims
C01facthigh confidence
AppLovin sells one thing: fees paid by advertisers whose campaigns its Axon AI engine bids and clears, and AppLovin Ads is the vast majority of that revenue.
C02facthigh confidence
The advertiser sets a return-on-ad-spend target and AppLovin's model decides which impression to buy at what price; revenue is what is charged less what is paid to the publisher.
C04facthigh confidence
Alphabet's own filing separates the half of the market AppLovin sells into: performance advertising connects advertisers to users while driving measurable results, against brand advertising, which builds awareness.
C05facthigh confidence
Unity's own annual report describes the identical product: monetization solutions letting publishers sell inventory to advertisers, with Unity retaining the amount from a transaction facilitated through its auction and mediation platform.
C06facthigh confidence
Cost of revenue was $225,801 thousand on $1,923,686 thousand of Q2 2026 revenue - 11.7% - and rose $70.7 million year on year, $52.1 million of it network infrastructure.
How it makes money
It is paid for each advertisement it places. No advertiser is under contract, so every dollar has to be won again at the next auction. What brings advertisers back is results, and the results are what the auction's model produces.
The evidence under this - 3 claims
C01facthigh confidence
AppLovin sells one thing: fees paid by advertisers whose campaigns its Axon AI engine bids and clears, and AppLovin Ads is the vast majority of that revenue.
C19facthigh confidence
AppLovin has no contracted advertiser demand: its own risk factors list 'risks related to not having long-term agreements with our clients', and every dollar of AppLovin Ads revenue is re-won at each auction.
C20facthigh confidence
AppLovin committed in August 2024 to spend a minimum of $1.3 billion with a cloud computing provider over three years and had paid $953.5 million of it by 30 June 2026.
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Advertising, United States users | 51.4% | Fees from advertisers whose campaigns Axon bids and clears, net of what publishers are paid, for ads shown to users in the United States. |
| Advertising, rest of the world | 48.6% | The same auction fees for ads shown to users outside the United States. |
| Other services (Adjust measurement, Wurl) | not material, per the 10-Q | Measurement and streaming-TV tools sold alongside the auction; the filing says their revenue is not material and gives no figure. |
Who runs it
Adam Foroughi co-founded AppLovin and has run it since 2011; Matthew Stumpf has run finance since 2024, and a new technology chief took over in July. Insiders own about an eighth of the company and, through extra-vote shares, control it. Executives get no cash bonus: their pay is shares that vest within a year, sized at the board's judgement rather than against any target.
The evidence under this - 5 claims
C63facthigh confidence
Adam Foroughi co-founded AppLovin and has been chief executive and a director since December 2011, and its chairperson from March 2021 to April 2026; Matthew Stumpf, at the company since 2020 after fourteen years at PwC, has been chief financial officer since January 2024.
C53facthigh confidence
AppLovin's CTO since 2020 stepped down on 1 July 2026, the day after the quarter the model improvement ran light, becoming a non-executive Distinguished Engineer; the successor, who led Axon 2.0, had run engineering and product since November 2025.
sourceAppLovin Form 8-K, Item 5.02, executive transitions, with…primaryAppLovin 2026 proxy statement, Form DEF 14Aprimary
against this
- The change was announced on 7 April, three months ahead, and the successor had already run the engineering and product teams for seven months, so the title moved on 1 July but the person running the work did not.
- A transition in the quarter the research output slipped is a coincidence of timing until something connects them, and management gave no such connection on the call; it is recorded because the ledger's timing account had never been read against it.
C61inferencehigh confidence
Directors and executive officers held 14,437,484 Class A and 28,634,575 Class B shares at 31 March 2026 - about 12.8% of the shares and 64.0% of the votes; the chief executive alone held about 9.0% of the shares, and no insider holds more of the company.
against this
- The chief executive sells about 100,000 shares each window at his own discretion and no insider bought in the year (C41), so the stake is shrinking slowly rather than growing.
- Of the other large insider holdings, a director's 6,963,289 Class A shares include 550,000 pledged under a prepaid variable forward due in May 2027.
C45facthigh confidence
Two people control the company: the co-founder and chief executive and one director hold all Class B stock at 20 votes a share, and shares under their voting agreement carried 66.9% of total voting power at 31 March 2026, 61.6% the chief executive's alone.
sourceAppLovin 2026 proxy statement, Form DEF 14AprimaryAppLovin Form 10-K, fiscal year ended 31 December 2025primary
- 2026-09-23editedRe-sourced from the 2026 proxy, which gives the voting-agreement percentage at 31 March 2026 rather than the annual report's year-end figure.
C62facthigh confidence
Executive pay has no annual cash bonus and a salary capped at $400,000 since before the 2021 listing; all variable pay is equity, granted in October 2025 as equal-sized RSUs vesting over one year, with no performance units granted in 2024 or 2025 and no company-selected financial measure.
against this
- The committee states it sizes each year's grant on the past year's company and individual performance, so the rewards are tied to results, but by discretion rather than to any stated metric.
- The 2023 performance units, which paid on the share price alone, were fully earned in 2024 and created most of the executives' wealth, so the stake already held matters more than the annual grant.
Why it wins, and who could beat it
Why customers come back
AppLovin sits on both sides of the same auction, so every advertiser's results train the model that serves all the others: the more it places, the better it places.
The evidence comes from outside the company. AppsFlyer, which measures app installs for its own clients, ranked it first in North America and Western Europe in December 2025, and it earns the highest margin of the five rivals its annual report names.
The evidence under this - 4 claims
C05facthigh confidence
Unity's own annual report describes the identical product: monetization solutions letting publishers sell inventory to advertisers, with Unity retaining the amount from a transaction facilitated through its auction and mediation platform.
C09facthigh confidence
Against the four peers its own 10-K names and whose perimeter is comparable, AppLovin's trailing operating margin of 77.4% is the highest, 39 points above Meta at 38.1%.
C56factmedium confidence
AppsFlyer, an attribution firm that competes with AppLovin's Adjust, ranked AppLovin second and third in gaming for the first half of 2024, and in December 2025 had it leading in tier-one North America and Western Europe and narrowing the gap on Apple Ads.
sourceAppsFlyer Performance Index, 2025 edition, press releaseprimaryAppsFlyer Performance Index, 17th edition, press releaseprimary
against this
- The index ranks media sources by installs AppsFlyer's own clients attribute to them, ordinally and with a weighting it does not publish, so it says where AppLovin stands and not by how much.
- Its latest edition predates the June 2026 quarter, so it describes the moat before the stall and cannot say whether the stall moved it.
C12facthigh confidence
Unity's 10-K names AppLovin, Voodoo, Moloco and Digital Turbine as the companies it competes against - naming the private attackers AppLovin's own filing leaves as 'various private companies'.
The threat
Unity rebuilt its own model. Its advertising business has sped up every quarter since, reaching 35% growth in June. That quarter it also grew faster than AppLovin from one quarter to the next, although from a base a fifth the size.
The evidence under this - 3 claims
C13facthigh confidence
Unity rolled out Unity Vector, a new machine-learning model, in its Grow Solutions business in the first quarter of 2025, and states that failing to keep improving it would harm the business.
C10facthigh confidence
Unity's Grow Solutions revenue was $1,228,239 thousand in fiscal 2025 against $1,199,289 thousand in 2024 - 2.4% growth in the year it rolled out Unity Vector, its new machine-learning advertising model.
against this
- Unity's own MD&A attributes the increase to the Vector migration and says the IronSource Ad Network declined, so the Vector-exposed network grew faster than 2.4% and a second, shrinking network masked it.
- Unity Vector shipped in the first quarter of 2025, so a full-year comparison contains three quarters of it at most and the exit rate is not readable from an annual figure.
- The year is no longer the reading: the same segment grew 35.4% year on year by June 2026 and outgrew AppLovin sequentially in that quarter (C58), so 2.4% measured the model's first year rather than what it produces.
- 2026-09-23editedThe counter-evidence said the Vector-exposed network was masked by a shrinking one; Unity's 2026 quarters now show the masking lifting, which belongs beside the annual figure.
C58inferencehigh confidence
Unity's Grow Solutions grew 6.4%, 10.9%, 23.5% and 35.4% year on year in the four quarters to June 2026, to $388,850 thousand, crediting Unity Vector; in the June quarter it grew 10.6% sequentially while AppLovin grew 4.4%.
sourceUnity Software Form 10-Q, quarter ended 30 June 2026primaryUnity Software Form 10-Q, quarter ended 30 September 2025primaryUnity Software Form 10-K, fiscal year ended 31 December 2025primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- AppLovin's revenue is 4.95 times Unity's whole segment and grew faster year on year in every one of the four quarters - 52.8% against 35.4% in the last.
- One quarter is one quarter: in the first quarter of 2026 Unity grew 4.0% sequentially against AppLovin's 11.1%, so the June reading may be Unity's own lumpiness as much as a shift.
The rivals: U, GOOGL, GOOGL, META, APPS, AMZN, and 1 more that is private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| U | Unity Software (Unity Vector)Priced against it: Yes | AppLovin Ads for mobile-game advertisers, bid by Axon | On sale | Yes |
| GOOGL | Alphabet (Performance Max)Priced against it: Yes | AppLovin Ads for web and e-commerce advertisers, the consumer vertical | On sale | Yes |
| Alphabet (AdMob mediation) | MAX, the publisher-side in-app bidding auction | On sale | ||
| META | Meta (self-service performance ads)Priced against it: Yes | AppLovin Ads for web and e-commerce advertisers, the consumer vertical | On sale | Yes |
| APPS | Digital TurbinePriced against it: Yes | AppLovin Ads for mobile-game advertisers, bid by Axon | On sale | Yes |
| Private | MolocoPriced against it: Private | AppLovin Ads for mobile-game advertisers, bid by Axon | On sale | Private |
| AMZN | Amazon.comPriced against it: Shown, not counted | - | On sale | Shown, not counted |
How big it could get
How big this could get
Americans spent $162B on advertising bought by automatic auction in 2025, up 20.5%, while advertising sold by negotiation shrank. AppLovin's American sales are 1.7% of that pool. Nobody measures its share of phone-game advertising itself.
The evidence under this - 3 claims
C15facthigh confidence
US internet advertising revenue was $294.6 billion in 2025, up 13.9%; the $162.4 billion cleared programmatically grew 20.5% while non-programmatic placements excluding search fell 13.9% to $18.0 billion.
C16facthigh confidence
US commerce media advertising revenue reached $63.4 billion in 2025, up 18.0% - a deceleration from 23.0% growth the year before.
C18inferencemedium confidence
AppLovin's FY2025 United States revenue of $2,827,248 thousand is 1.74% of the $162.4 billion of US advertising revenue cleared programmatically that year.
sourceAppLovin Form 10-K, fiscal year ended 31 December 2025primaryIAB/PwC Internet Advertising Revenue Report, full-year 2025…primary
against this
- AppLovin's geographic split is by user location, not by where the advertiser buys, so a US impression sold to a foreign advertiser counts here and may not count in the IAB survey the same way.
- IAB's programmatic line excludes search but includes social, and Meta and Alphabet own most of the latter, so the pond an independent bidder can actually reach is smaller than 1.74% implies.
The results
The numbers
Is it growing?Fast
Sales grew 70% in the year to Dec 2025, slower than the 75% in the year before.
Is it profitable?Strong
Operating margin is 76%, up from −1.7% in the year to Dec 2022.
Does growth pay off?Pays off
Running costs take 12¢ of each sales dollar, down from 59¢ in the year to Dec 2021.
Does it earn good returns?Strong
Each dollar put into the business earns 70% a year after tax, up from 3.4% in the year to Dec 2021.
Is it making cash?Strong
The business brought in $3.97B of cash from operations, 72% of sales, before capital spending, which the filer does not tag.
Is the balance sheet safe?Light debt
It owes $1.03B more than the cash it holds, about 0.3 years of the cash its operations bring in.
Are shares being diluted?Stable
The share count barely moved in 4.0 years, and pay in shares costs 3.9% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | |
|---|---|---|---|---|---|
| Size and growth | |||||
| Sales | $2.79B | $2.82B | $1.84B | $3.22B | $5.48B |
| Sales growth | +92% | +0.9% | −35% | +75% | +70% |
| Profit | |||||
| Gross margin | 65% | 55% | 81% | 84% | 88% |
| Operating margin | 5.4% | −1.7% | 42% | 59% | 76% |
| Net margin | 1.3% | −6.8% | 19% | 49% | 61% |
| What growth costs, per sales dollar | |||||
| Research and development | 13% | 18% | 18% | 12% | 4.1% |
| Sales and marketing | 40% | 33% | 12% | 7.8% | 3.7% |
| General and admin | 5.7% | 6.4% | 8.2% | 5.1% | 4.3% |
| Pay in shares | 4.8% | 6.8% | 20% | 12% | 3.9% |
| Capital spending | 0.0% | 0.0% | 0.2% | 0.1% | - |
| Cash | |||||
| Cash from operations | $362M | $413M | $1.06B | $2.10B | $3.97B |
| Free cash flow | $360M | $412M | $1.06B | $2.09B | - |
| Balance sheet | |||||
| Cash and short-term investments | $1.52B | $1.08B | $502M | $697M | $2.49B |
| Debt | $3.25B | $3.25B | $3.34B | $3.51B | $3.51B |
| Net cash | −$1.73B | −$2.16B | −$2.83B | −$2.81B | −$1.03B |
| Returns and shares | |||||
| Return on capital | 3.4% | - | 14% | 41% | 70% |
| Shares (diluted) | 343M | 372M | 363M | 348M | 342M |
What growth costs
Growth has made AppLovin far more profitable: sales nearly tripled in two years while its costs barely moved, so each new dollar of sales costs much less to win. Computing power is the one cost that rises with sales. Part of the drop was smaller stock awards, and those are climbing again.
The evidence under this - 3 claims
C59facthigh confidence
Total costs and expenses were $1,069,351 thousand, $1,313,102 thousand and $1,328,803 thousand in 2023, 2024 and 2025 against revenue of $1,841,762 thousand, $3,224,058 thousand and $5,480,717 thousand - 58.1%, 40.7% and 24.2% of revenue; research and development fell from 18.1% to 4.1%, sales and marketing from 12.4% to 3.7%, cost of revenue from 19.4% to 12.1%.
against this
- Most of 2025's 40% fall in research and development was a $151.1 million drop in stock-based payroll costs, not fewer engineers doing more, and research stock pay quadrupled year on year in the June 2026 quarter (C34).
- The three years are the advertising business restated without the Apps business; the numbers card's 2021 and 2022 columns still carry the games and are a different company.
C60facthigh confidence
Stock-based compensation in continuing operations was $342,551 thousand, $357,431 thousand and $207,958 thousand in 2023, 2024 and 2025 - 18.6%, 11.1% and 3.8% of revenue - of which research and development carried $216,236 thousand, $229,577 thousand and $114,463 thousand.
C26management claimmedium confidence
The CFO states AppLovin spends about $0.10 of compute for each incremental dollar of revenue and expects no departure from that level.
against this
- The same call says a materially larger model would be pushed live regardless, so the ratio is a policy the company can abandon in a quarter rather than a constraint on it.
- Cost of revenue rose 46% year on year against revenue growth of 53%, which is consistent with the ratio and does not isolate compute from the rest of network infrastructure.
The last report (5 Aug)
Quarter ended June 2026, filed figures.
53%
Sales growth
against the same quarter a year earlier
4.4%
Growth from the quarter before
against 11.1% the quarter before that
78¢
Operating profit
per dollar of sales
On the surface it was another record. Sales grew 53% and earnings per share of $3.76 matched the forecast exactly, with no adjustments at all.
The problem was direction. Sales grew only 4.4% from the quarter before and landed below the middle of the company's own guidance, the first time that has happened since it listed. The shares fell 19.7% the next day.
What it did to the thesisThis is the quarter the whole case is about. Growth slowed sharply and landed below the company's own forecast, which the market read as the end of the story. The thesis reads it as mostly timing, hence the Attractive view, but low conviction admits the next quarter must prove it.
The quarter in detail
Sales growth from the previous quarter, filed figures. The company has restated its own history, so a quarter either side of the restatement may describe a different business; n/c marks one that does not compare.
Earnings per share $3.76 against $3.76 expected; $3.76 as filed, with no adjustments.
The evidence under this - 4 claims
C21facthigh confidence
Q2 2026 revenue of $1,923,686 thousand grew 52.8% year on year but only 4.4% sequentially, against 11.1% the quarter before and 18.0% the quarter before that.
C30facthigh confidence
In the three most recent prints carrying a filed diluted figure, adjusted earnings per share equals filed GAAP diluted earnings per share to the cent - $3.76, $3.56 and $2.45, with add-backs of zero.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- Adjusted EBITDA is a different matter and does carry add-backs: $1,613,823 thousand for Q2 2026 against $1,494,277 thousand of income from operations, the difference being mostly $85,783 thousand of stock compensation.
- The August 2025 print had adjusted EPS of $2.26 against a filed $2.39, so the two series have not always agreed - though that gap ran in the shareholder's favour.
C50facthigh confidence
The June quarter was guided in May to revenue of $1,915 to $1,945 million and adjusted EBITDA of $1,615 to $1,645 million; revenue of $1,923.7 million landed below the midpoint and adjusted EBITDA of $1,613.8 million landed below the floor.
C39inferencemedium confidence
Three brokers cut the stock within a week of the print and the shares fell 19.7% the day after it, 6.0% five sessions later - the price move is a repricing of the growth rate, not of the quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- The July fall came first and on a different cause - an analyst note flagging cooled e-commerce sign-ups in June - so the drawdown is at least two events and only one of them is the print.
- A stock priced at 54 times earnings does not need a large miss to lose a fifth of its value; the magnitude is a function of where the multiple started.
What management did with the money
It bought back shares, cutting the count from 338 to 335 million in six months while spending 8.9 times what it paid staff in stock. Borrowings are covered almost entirely by cash.
Insiders in 2026Q2: 9 purchases and 180 sales, from their Form 4 filings.
The evidence under this - 2 claims
C33facthigh confidence
Shares outstanding fell from 338 million at the end of 2025 to 335 million at 30 June 2026; the company retired 3,275,252 shares for $1.5 billion in the half, and in 2025 repurchased and retired 5,511,519 shares for $2.2 billion, or 6.4 million for $2.58 billion counting shares withheld for employees' taxes.
sourceAppLovin Form 10-Q, quarter ended 30 June 2026primaryAppLovin fourth quarter and full year 2025 results release,…primaryAppLovin Form 10-K, fiscal year ended 31 December 2025primary
- 2026-10-04editedThe 2025 figure said 6.4 million shares were retired for $2.58 billion; the release (S04) says 'repurchased and withheld', and the 10-K equity note (S01) gives 5,511,519 shares repurchased and retired for $2.2 billion, the rest being shares withheld for employees' taxes.
C31inferencehigh confidence
Cash was $3,053,306 thousand against total debt of $3,515,072 thousand at 30 June 2026: net debt of $461,766 thousand, about 0.08 times trailing adjusted EBITDA of roughly $5.7 billion.
sourceAppLovin Form 10-Q, quarter ended 30 June 2026primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primaryAppLovin fourth quarter and full year 2025 results release,…primary
against this
- The denominator is adjusted EBITDA, which adds back $169,252 thousand of stock compensation in the half; on income from operations the same net debt is a higher multiple, though still under 0.1 times.
- The trailing figure spans the Apps divestiture, so the FY2025 component includes a company AppLovin no longer owns - the same perimeter problem the pack refuses to compute a revenue figure across.
The price, and where we disagree
In short. Cheap. The price assumes AppLovin's growth is nearly over. If November's results land inside the promised range, that bet loses; if they miss, the market was right.
Share price
$277.04
close,
Price for each $1 of spare cash
24.5
rivals' middle 57.7
Spare cash as a share of the price
4.1%
its 6-year normal 1.8%
Two things make this the moment to look. First, the shares have halved for the second time in six months. After the last halving, in February, the next results sent them back up within four months.
Second, the question is about to be answered. In November AppLovin reports the September quarter, the first real test of management's explanation, against a promise of $2.06B to $2.09B of sales.
Underneath, the market keeps moving its way: advertising bought by automatic auction grew 20.5% in 2025, while the negotiated kind shrank 13.9%.
The evidence under this - 4 claims
C54facthigh confidence
The shares fell 50.0% from $733.60 on 22 December 2025 to $366.91 on 12 February 2026, 22.4% of it in the session after a fourth-quarter print that beat, on objections about Meta, Google and a short report; after the first-quarter print they closed at $613.70 on 1 June.
sourceFMP end-of-day price history for APP, 15 April 2021 to 22…primaryThe Motley Fool, 'Why AppLovin Stock Fell 24% in the First…secondaryAppLovin first quarter 2026 results release, Exhibit 99.1 to…primary
against this
- The coverage names three objections at once and a software sell-off beside them, so the February fall cannot be pinned on the one objection this thesis is about.
- A recovery says the market changed its mind after the next print; it does not say the objection was wrong, and the same shares then halved again from 1 June.
C24facthigh confidence
Third-quarter 2026 guidance is revenue of $2,055 to $2,085 million and adjusted EBITDA of $1,710 to $1,740 million, and management states it assumes no model releases that have not already been deployed.
C22management claimmedium confidence
Management attributes the sequential slowdown to timing in its own research output: the pace of meaningful model improvement was lighter than normal and the next step-up landed just after quarter end.
against this
- It is the explanation most favourable to the company, offered by the company, for the first guidance-midpoint miss in its public history, and no filed document corroborates the timing of a model release.
- Bank of America had already flagged in July that new e-commerce sign-ups cooled in June, which is a demand-side account of the same quarter and is not what management describes.
- A complaint filed in September 2026 puts a competing account of the same quarter on the record: it alleges the constancy of the model improvement was overstated throughout a class period ending on the day of that print. It is an allegation and decides nothing, and it is the second named account of this quarter that is not management's.
- Unity's advertising segment, running on its rival model, grew 10.6% sequentially in the same quarter against AppLovin's 4.4% - a filed figure for a competitive account of the quarter, where the others are an analyst note and a pleading.
- The chief technology officer since 2020 stepped down on 1 July, the day the quarter closed; the handover was announced in April and the successor already ran engineering, so it explains nothing by itself, but management's account of the quarter does not mention it.
- 2026-09-21editedThe counter-evidence held one outside account of the quarter. A second class action now alleges the opposite of what management said about the model cadence, which belongs beside it.
- 2026-09-23editedTwo facts about the same quarter had not been read against the claim: a filed rival figure that grew faster, and a change of technical leadership on the day it closed.
C15facthigh confidence
US internet advertising revenue was $294.6 billion in 2025, up 13.9%; the $162.4 billion cleared programmatically grew 20.5% while non-programmatic placements excluding search fell 13.9% to $18.0 billion.
The betAsymmetric
If it works
2.2×
from today's value, at the stated multiple
If it fails
keep 100%
of today's market value: the rest of the company, less its debt
How we read the multiple
Buyers pay about 24.5 times the spare cash AppLovin produced last year, a 4.1% yearly return against 1.8% normally. Its rivals cost about 58 times, so it is the cheapest of them.
That makes sense only if growth is nearly over. Yet sales grew 47.6% last year, and the company expects 46% to 48% growth in the September quarter even with no new model releases.
For the price to be wrong, the improvements only have to keep arriving.
Against its own history. Far cheaper than this company has ever been, while the business is bigger and more profitable than it was, so the cheapness is not decay.
The evidence under this - 4 claims
C36facthigh confidence
The shares trade at 25.2 times trailing earnings against a six-year median of 54.0, and at a 4.1% free cash flow yield against a six-year median of 1.8%; the company's own releases put free cash flow at $3.95 billion in 2025 and $863.3 million in the June quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primaryAppLovin fourth quarter and full year 2025 results release,…primary
- 2026-09-23editedMultiples refreshed with the pack on 23 September 2026, at a price 10.1% higher.
C37facthigh confidence
At 24.5 times free cash flow AppLovin is the cheapest of the five companies its own annual report names, against a cohort median of 57.7 - Unity 35.0, Meta 45.8, Digital Turbine 69.6, Alphabet 79.1. Its own free cash flow is filed: $3.95 billion in 2025.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin fourth quarter and full year 2025 results release,…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primary
against this
- On sales it is the dearest of the same five at 16.2 times against a median of 8.75, so the answer inverts with the metric and the choice of metric is doing the work.
- A free cash flow multiple flatters a company whose free cash flow has just grown 58% and would look very different against the flat figure the bear case expects.
- 2026-09-23editedCohort refreshed with the pack on 23 September 2026: AppLovin still ranks cheapest, and the peer figures moved.
C38inferencemedium confidence
The price pays 24.5 times free cash flow for a business that grew revenue 47.6% over the trailing year, earns 63.8% on invested capital and carries net debt of 0.08 times EBITDA - it assumes the growth is close to over.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primary
against this
- Free cash flow is a trailing figure struck before the compute step-up management has already put into guidance, so the denominator may be about to stop growing at the rate that makes the multiple look low.
- A cheap multiple on a business whose central input is unpredictable research output is exactly what a value trap looks like from the outside, and the last six quarters of multiple compression are the market saying so.
- 2026-09-23editedRefreshed with the pack on 23 September 2026.
C24facthigh confidence
Third-quarter 2026 guidance is revenue of $2,055 to $2,085 million and adjusted EBITDA of $1,710 to $1,740 million, and management states it assumes no model releases that have not already been deployed.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on price for each $1 of spare cash, cheapest first:
| Ticker | Company | price for each $1 of spare cash |
|---|---|---|
| APP | AppLovin | 24.5× |
| U | Unity Software | 35.0× |
| META | Meta Platforms | 45.8× |
| APPS | Digital Turbine | 69.6× |
| GOOGL | Alphabet | 79.1× |
| AMZN | Amazon.com not in the median | not meaningful |
The middle of the other 4 is 57.7. Amazon.com, Inc. competes with it in one product line inside a much bigger company, so its price says little about this business and it is not counted in the middle. AppLovin Corporation is cheapest of 5.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
The cheapest of the rivals its own report names, on the measure this thesis prices it by. That position is its own, not the group's.
Private or foreign rivals that cannot be priced: Moloco, Voodoo, various private companies.
The bet: what it pays if it works
The bet is on adverts for online shops, opened to every advertiser only in June and not reported apart from games, so the game adverts and the publisher auction are counted separately, at the price per dollar of spare cash this thesis would call neutral.
Counted that way, the game business alone is worth more than the whole company today. If shops sign up as management hopes, they add less than double on top; if they stay with Meta and Google, nothing is lost on them.
| AppLovin Ads for web and e-commerce advertisers, the consumer vertical revenue in two to three years | Consumer advertisers needed | At the stated multiple |
|---|---|---|
| $3.00B (the big win) | about 85.7K at 50% kept | 2.2× |
| $6.00B | about 171.4K at 50% kept | 3.0× |
To pay 2× (a Fair bet) it would need about $2.01B of AppLovin Ads for web and e-commerce advertisers, the consumer vertical revenue (about 57.3K consumer advertisers at 50% kept); to pay 5× (a Cheap bet), about $14.15B of AppLovin Ads for web and e-commerce advertisers, the consumer vertical revenue (about 404.2K consumer advertisers at 50% kept), at the same multiple.
Each row is AppLovin Ads for web and e-commerce advertisers, the consumer vertical alone, plus the rest of the company (Gaming ads and MAX) counted at about $140.00B.
The steps it has to take
- Done Consumer platform opened to all advertisers
- Done Consumer spend above its December-quarter peak
- In progress Self-serve advertisers beyond the mid-market due 2027-06
- Not started Consumer sales quantified in a filing or release due 2027-12
The detail behind the label
Rating
Fair
the big win's payoff against the 2× and 5× lines
Cheap bet at about
$124.43
55% below the price (close, )
The option
free
what AppLovin Ads for web and e-commerce advertisers, the consumer vertical is worth if it works, and the price pays nothing for it
The evidence under this - 5 claims
C47management claimlow confidence
Advertisers on the consumer platform are not near their spending ceiling, according to management, and the platform was opened to the public only in late June 2026.
against this
- The claim is the company reporting what its own customers tell it, on the one subject where a seller has the strongest reason to be optimistic, and no third-party spend data corroborates it.
- An analyst put the same question the other way - whether the merchants saying this are a positively biased sample - and management did not produce a figure in reply.
C65management claimlow confidence
Management said in May 2026 that the consumer business had passed a $1 billion run rate of spend a year earlier and is 'much bigger now', with April larger than any month of the December quarter; in February it declined to report e-commerce apart from gaming.
sourceAppLovin first quarter 2026 earnings call (FMP), the remarks…primaryAppLovin fourth quarter 2025 earnings call (FMP), the answer…primary
against this
A run rate of advertiser spend is not revenue, which is spend less what publishers are paid, and 'much bigger' is not a figure; nothing filed corroborates either.
C66management claimlow confidence
Management projected in May 2026 'well over $70,000 a year' of spend from each new consumer advertiser, and said 100,000 such advertisers signed in a year would bring roughly $7 billion of first-year advertising spend.
against this
The figure comes from the first, referral-chosen cohort; management itself says small shops struggle to reach their goals (C46), so later advertisers may spend less.
C67estimatelow confidence
Taking AppLovin's trailing free cash flow at about $4.5 billion (24.5 times on a $110.4 billion market value) and removing roughly $0.75 billion of consumer sales at the company's own cash margin of about 66 cents per sales dollar leaves about $4.0 billion of free cash flow for gaming ads and MAX.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin second quarter 2026 results release, Exhibit 99.1…primary
against this
The consumer split is not disclosed (C65); if consumer sales are larger than assumed, the core is smaller and the option less free.
C46management claimhigh confidence
The consumer platform's own gaps are stated by management: creative is the biggest hurdle, a high-quality long-form video advertisement cannot yet be generated out of the box, and mid-market rather than the long tail is where it works.
against this
- Management said the same thing about creative a quarter earlier, when the self-service funnel converted 57% of qualified leads and lack of creative was named as the leak - so the gap has been open for at least two quarters with no shipped fix.
- Conceding that the long tail does not work today is also the most disarming way to present it, and no third party has measured how much of the addressable merchant base the mid-market actually is.
- The gap management concedes is what a September 2026 complaint alleges was understated: that the generative video creative feature faced significant development delays making release on the stated timeline unlikely. Conceding a hurdle and being accused of understating it are different claims about the same feature, and only the first is on the record as fact.
- 2026-09-21editedA new complaint alleges the creative delay this claim records as a conceded hurdle was understated, which is the counter-account the field was missing.
Where we disagree with the market
The market believes
The model improvements have run out, and the promised 30% a year was a story.
We believe
June was mostly timing: the marketplace still grew and consumer spending set a record.
The market is reacting to what it can see: shares far below their December high, three brokers cutting within a week, and fewer analysts positive than in May.
The case against us is real, and we found it by reading both companies' filings quarter by quarter. Unity grew 10.6% from the quarter before while AppLovin grew 4.4%, and no insider bought after the fall.
What analysts say, and the evidence
Analysts positive
79%
of 34, 1 Sep
Four months earlier
88%
of 32, 1 May
The evidence under this - 5 claims
C22management claimmedium confidence
Management attributes the sequential slowdown to timing in its own research output: the pace of meaningful model improvement was lighter than normal and the next step-up landed just after quarter end.
against this
- It is the explanation most favourable to the company, offered by the company, for the first guidance-midpoint miss in its public history, and no filed document corroborates the timing of a model release.
- Bank of America had already flagged in July that new e-commerce sign-ups cooled in June, which is a demand-side account of the same quarter and is not what management describes.
- A complaint filed in September 2026 puts a competing account of the same quarter on the record: it alleges the constancy of the model improvement was overstated throughout a class period ending on the day of that print. It is an allegation and decides nothing, and it is the second named account of this quarter that is not management's.
- Unity's advertising segment, running on its rival model, grew 10.6% sequentially in the same quarter against AppLovin's 4.4% - a filed figure for a competitive account of the quarter, where the others are an analyst note and a pleading.
- The chief technology officer since 2020 stepped down on 1 July, the day the quarter closed; the handover was announced in April and the successor already ran engineering, so it explains nothing by itself, but management's account of the quarter does not mention it.
- 2026-09-21editedThe counter-evidence held one outside account of the quarter. A second class action now alleges the opposite of what management said about the model cadence, which belongs beside it.
- 2026-09-23editedTwo facts about the same quarter had not been read against the claim: a filed rival figure that grew faster, and a change of technical leadership on the day it closed.
C23management claimmedium confidence
Two checkable figures in the same quarter argue against a demand or competitive explanation: MAX publisher earnings grew double digits sequentially, and consumer advertiser spend finished 28% above Q4 2025, the seasonal peak.
against this
- Both figures are stated on the call and appear in no filing, so neither can be checked against a filed number; the CFO had to correct the CEO from 26% to 28% live on the call.
- A publisher marketplace growing is consistent with AppLovin losing share inside a growing market, since the statistic measures the pond rather than the catch.
- The competitive explanation now has a filed figure behind it: Unity's segment grew 10.6% sequentially in the quarter these two figures describe (C58).
- 2026-09-23editedThe claim argued its two figures tell against a competitive explanation; a named rival's filing for the same quarter now argues for one.
C39inferencemedium confidence
Three brokers cut the stock within a week of the print and the shares fell 19.7% the day after it, 6.0% five sessions later - the price move is a repricing of the growth rate, not of the quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- The July fall came first and on a different cause - an analyst note flagging cooled e-commerce sign-ups in June - so the drawdown is at least two events and only one of them is the print.
- A stock priced at 54 times earnings does not need a large miss to lose a fifth of its value; the magnitude is a function of where the multiple started.
C58inferencehigh confidence
Unity's Grow Solutions grew 6.4%, 10.9%, 23.5% and 35.4% year on year in the four quarters to June 2026, to $388,850 thousand, crediting Unity Vector; in the June quarter it grew 10.6% sequentially while AppLovin grew 4.4%.
sourceUnity Software Form 10-Q, quarter ended 30 June 2026primaryUnity Software Form 10-Q, quarter ended 30 September 2025primaryUnity Software Form 10-K, fiscal year ended 31 December 2025primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- AppLovin's revenue is 4.95 times Unity's whole segment and grew faster year on year in every one of the four quarters - 52.8% against 35.4% in the last.
- One quarter is one quarter: in the first quarter of 2026 Unity grew 4.0% sequentially against AppLovin's 11.1%, so the June reading may be Unity's own lumpiness as much as a shift.
C35facthigh confidence
At $328.73 on 22 September 2026 the shares are 55.9% below a 52-week high of $745.61 and 10.5% above the 52-week low, having risen 10.1% over twenty sessions from that low.
- 2026-09-23editedPrice refreshed with the pack on 23 September 2026.
What could go wrong, and how we would know
The risks
- Demand, not timing. An outside analyst flagged cooling e-commerce sign-ups a month before the results.
- A rival catching up. Unity's advertising business outgrew AppLovin's from one quarter to the next in June.
- Growth nobody can schedule. The engine is research output, and management said as much on the call.
- Two shareholder lawsuits. The newer one says the company overstated how steadily its models improved.
- A gap in the new business. The e-commerce side still cannot make the video advertisements its smaller customers need.
The evidence under this - 6 claims
C22management claimmedium confidence
Management attributes the sequential slowdown to timing in its own research output: the pace of meaningful model improvement was lighter than normal and the next step-up landed just after quarter end.
against this
- It is the explanation most favourable to the company, offered by the company, for the first guidance-midpoint miss in its public history, and no filed document corroborates the timing of a model release.
- Bank of America had already flagged in July that new e-commerce sign-ups cooled in June, which is a demand-side account of the same quarter and is not what management describes.
- A complaint filed in September 2026 puts a competing account of the same quarter on the record: it alleges the constancy of the model improvement was overstated throughout a class period ending on the day of that print. It is an allegation and decides nothing, and it is the second named account of this quarter that is not management's.
- Unity's advertising segment, running on its rival model, grew 10.6% sequentially in the same quarter against AppLovin's 4.4% - a filed figure for a competitive account of the quarter, where the others are an analyst note and a pleading.
- The chief technology officer since 2020 stepped down on 1 July, the day the quarter closed; the handover was announced in April and the successor already ran engineering, so it explains nothing by itself, but management's account of the quarter does not mention it.
- 2026-09-21editedThe counter-evidence held one outside account of the quarter. A second class action now alleges the opposite of what management said about the model cadence, which belongs beside it.
- 2026-09-23editedTwo facts about the same quarter had not been read against the claim: a filed rival figure that grew faster, and a change of technical leadership on the day it closed.
C39inferencemedium confidence
Three brokers cut the stock within a week of the print and the shares fell 19.7% the day after it, 6.0% five sessions later - the price move is a repricing of the growth rate, not of the quarter.
sourceResearch pack for APP, frozen beside this thesis (refreshed…primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- The July fall came first and on a different cause - an analyst note flagging cooled e-commerce sign-ups in June - so the drawdown is at least two events and only one of them is the print.
- A stock priced at 54 times earnings does not need a large miss to lose a fifth of its value; the magnitude is a function of where the multiple started.
C25management claimlow confidence
Management states it believes the business can compound at roughly 30% annually over the longer term.
against this
- Nothing is contracted behind it - see the absence of any long-term client agreement - so it is an aspiration about an auction that is re-run every day.
- Three brokers cut the stock within a week of the print citing exactly this: diminished visibility into whether the 30% trajectory is sustainable.
C42facthigh confidence
Two securities class actions are live and they allege different things. The first, the Brownback action, alleges materially false statements about the advertising solutions and the financial growth; its motion to dismiss was fully briefed in February 2026 and remained undecided at 5 August 2026. The second was filed in September 2026 on the model-improvement and generative creative statements, and is the subject of C52.
sourceAppLovin Form 10-Q, quarter ended 30 June 2026primaryAppLovin Form 10-K, fiscal year ended 31 December 2025primaryRosen Law Firm notice: securities class action on behalf of…secondary
- 2026-09-21editedThe claim counted one class action. A second was filed in September 2026 on different allegations, so a claim naming one had become false.
C46management claimhigh confidence
The consumer platform's own gaps are stated by management: creative is the biggest hurdle, a high-quality long-form video advertisement cannot yet be generated out of the box, and mid-market rather than the long tail is where it works.
against this
- Management said the same thing about creative a quarter earlier, when the self-service funnel converted 57% of qualified leads and lack of creative was named as the leak - so the gap has been open for at least two quarters with no shipped fix.
- Conceding that the long tail does not work today is also the most disarming way to present it, and no third party has measured how much of the addressable merchant base the mid-market actually is.
- The gap management concedes is what a September 2026 complaint alleges was understated: that the generative video creative feature faced significant development delays making release on the stated timeline unlikely. Conceding a hurdle and being accused of understating it are different claims about the same feature, and only the first is on the record as fact.
- 2026-09-21editedA new complaint alleges the creative delay this claim records as a conceded hurdle was understated, which is the counter-account the field was missing.
C58inferencehigh confidence
Unity's Grow Solutions grew 6.4%, 10.9%, 23.5% and 35.4% year on year in the four quarters to June 2026, to $388,850 thousand, crediting Unity Vector; in the June quarter it grew 10.6% sequentially while AppLovin grew 4.4%.
sourceUnity Software Form 10-Q, quarter ended 30 June 2026primaryUnity Software Form 10-Q, quarter ended 30 September 2025primaryUnity Software Form 10-K, fiscal year ended 31 December 2025primaryAppLovin Form 10-Q, quarter ended 30 June 2026primary
against this
- AppLovin's revenue is 4.95 times Unity's whole segment and grew faster year on year in every one of the four quarters - 52.8% against 35.4% in the last.
- One quarter is one quarter: in the first quarter of 2026 Unity grew 4.0% sequentially against AppLovin's 11.1%, so the June reading may be Unity's own lumpiness as much as a shift.
What we still do not know
- AppLovin's share of phone-game advertising. Nobody publishes it, so its lead has to be read from a rival's accounts.
- How big the e-commerce business is. It is not reported separately, yet it carries much of the growth case.
The evidence under this - 3 claims
C11inferencemedium confidence
AppLovin's FY2025 revenue of $5,480,717 thousand is 4.46 times Unity's entire Grow Solutions segment, the largest listed business doing the same job.
sourceAppLovin Form 10-K, fiscal year ended 31 December 2025primaryUnity Software Form 10-K, fiscal year ended 31 December 2025primary
against this
- Meta and Alphabet each sell more performance advertising than both of them together, so 'largest' holds only among the companies whose whole business is bidding on somebody else's inventory.
- Moloco and Voodoo are private and file nothing, so the denominator of any share statement is unknowable from filings alone.
C12facthigh confidence
Unity's 10-K names AppLovin, Voodoo, Moloco and Digital Turbine as the companies it competes against - naming the private attackers AppLovin's own filing leaves as 'various private companies'.
C47management claimlow confidence
Advertisers on the consumer platform are not near their spending ceiling, according to management, and the platform was opened to the public only in late June 2026.
against this
- The claim is the company reporting what its own customers tell it, on the one subject where a seller has the strongest reason to be optimistic, and no third-party spend data corroborates it.
- An analyst put the same question the other way - whether the merchants saying this are a positively biased sample - and management did not produce a figure in reply.
What would change our view
- Attractive→NeutralThe multiple returns toward its own history
It needs:
- ○The price climbs to about 35 times the cash the business produces: still far below its own history, but the panic discount gone.about $395.77 today, 43% above the last close
- Attractive→UnattractiveThe stall turns out to be the model
It needs any of:
- ○A second quarter in a row of single-digit growth from the quarter before.checked at the next results, due in Nov 2026
- ○September-quarter sales below the $2.06B the company promised as its floor.checked at the next results, due in Nov 2026
- ○The accounting lawsuit survives the company's request to dismiss it.watched as it happens
- ○Unity's advertising business outgrows AppLovin's from one quarter to the next for a second quarter running.checked at the next results, due in Nov 2026
We would also be wrong if
- Operating profit falls below 70 cents in the dollar because computing costs rise faster than sales.
What is coming
- 4 November 2026 Q3 2026 results: Revenue against the $2,055-2,085 million guide and whether sequential growth returns above 7%, which tests the timing account of June's slowdown.
Every dated event ahead
- 4 November 2026 Q3 2026 results: Revenue against the $2,055-2,085 million guide and whether sequential growth returns above 7%, which tests the timing account of June's slowdown.
- 4 November 2026 Unity's Q3 2026 results: Whether Unity's advertising segment outgrows AppLovin sequentially a second quarter, the competitive reading of June against the timing one.
- 16 November 2026 Lead plaintiff deadline in the second securities class action: Who leads the 2026 class-period case; a ruling on the first case's long-briefed motion to dismiss may come at any time.
- December 2026 AppsFlyer Performance Index, next edition: Whether an outside measurer still ranks AppLovin first in tier-one markets after a year of Unity's acceleration.
- February 2027 Q4 2026 results and the FY2026 10-K: A full year of the advertising-only company, the first trailing figure free of the restatement, and the 2027 guide against the 30% compounding claim.
Since we wrote this
Written
25 Aug
Price the thesis read
$328.73
22 Sep
Latest close
$277.04
close, · -15.7% since
- 12 OctBet moveBet re-struck: Safe, little upside to AsymmetricRe-priced at the 9 October close; the rating moved from Expensive bet to Fair bet.
- 11 OctThesis reviewOnline shop ads declared an early line, with a bet on themThe owner asked every thesis to name the lines its story rides; the consumer business carries assumption A2 and is not reported apart.
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 23 SepThesis reviewConviction cut from Medium to LowUnity's advertising business, the rival the moat is measured against, has sped up every quarter and outgrew AppLovin in June, the quarter we had called timing.
- 21 SepNewsA second shareholder lawsuit. View heldIt alleges the company overstated how steadily its models improve. That is an allegation, not a finding, and November's results test the same question.
- 15 SepThesis reviewLegal disclosures re-read. View heldOne undecided court motion now holds up both the accounting lawsuit and the shareholder suits filed after it. Nothing here touches the business case.
- 25 AugThesis writtenAttractive, medium convictionThree weeks after June's results sent the shares to a 52-week low.
Also checked 6 times with nothing new to act on.
Sources 28
Where this came from
Sources - 28
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-19Item 1 on the four products and the named competitor set, the flywheel the company states in its own words, FY2025 statements and the geographic split, employee count, the senior notes and the revolver, the three-class share structure and the 67% voting agreement, and Item 3 on the securities and derivative litigationbacksC01C02C03C07C11C12C17C18C19C27C28C32C33C42C45C55C59C60C64
- S02primarycompany filing2026-08-05the Q2 2026 income statement and per-share figures, the Apps divestiture note, the cloud-computing purchase commitment, stock compensation by line, the H1 buyback, cash and debt, the Tripledot related-party revenue, and the litigation statusbacksC01C02C03C06C08C09C19C20C21C50C29C30C31C32C33C34C38C39C42C44C49C58C68
- S03primarycompany release2026-08-05the quarter's headline figures, free cash flow, the share count and the third-quarter guidance range
Revenue $1,924 million, up 53%; Adjusted EBITDA $1,614 million, up 58%; Free Cash Flow $863.3 million
backsC24C50C31C36C37C38C48C67 - S04primarycompany release2026-02-11the full-year 2025 revenue, net income, adjusted EBITDA and free cash flow, the year's buyback and the closing share count, and the first-quarter 2026 guidance the year was launched on
Revenue $1,658 $999 66% $5,481 $3,224 70% Net Income $1,102 $599 84%
backsC27C31C33C36C37C48 - S05primarytranscript2026-08-05management's account of what went wrong in the quarter, the MAX marketplace and consumer figures that appear in no filing, the compute-per-revenue-dollar ratio, the close of the SEC inquiry, and the mid-market sequencing of the consumer launch
Our pace of meaningful model improvement was lighter than normal during the quarter and the next step-up in model performance landed just after quarter end.
backsC22C23C24C25C26C43C46C47C48 - S06primarycompany filing2026-02-11Unity Software Form 10-K, fiscal year ended 31 December 2025 Unity Software Inc.the rival's own description of the same auction and mediation product, the date its competing machine-learning model shipped, the names of the private attackers AppLovin's own filing hides behind a class, and the segment revenue that says what the attack produced
Examples of other companies we compete against include AppLovin, Voodoo, Moloco, and Digital Turbine.
backsC05C09C10C11C12C13C58 - S07primarycompany filing2026-05-26Digital Turbine Form 10-K, fiscal year ended 31 March 2026 Digital Turbine, Inc.a second listed rival's own description of the same buyer and the same job, its segment split, and the near-identical wording of its Competition section
App Developers and other performance-focused advertisers execute mobile user acquisition campaigns for their apps and products on the Company's direct mobile app inventory.
backsC14 - S08primaryindustry data2026-04-30IAB/PwC Internet Advertising Revenue Report, full-year 2025 results Interactive Advertising Bureau and PricewaterhouseCoopersthe size and growth of the pond measured by somebody other than an advertising company: the US total, the programmatic share of it, the commerce-media line and the shrinking non-programmatic remainder
U.S. digital advertising revenues, which rose 13.9% year over year (YoY) in 2025 to a record $294.6 billion. ... Programmatic advertising revenue grew by $27.6 billion to a total of $162.4 billion in 2025, a YoY increase of 20.5%. ... non-programmatic placements (excluding search) declined 13.9% but still totalled $18.0 billion ... Commerce media advertising revenues increased by $9.7 billion to $63.4 billion in 2025, representing 18.0% year-over-year growth, down from 23.0% ... Search revenues maintained the largest market share at 38.8% but grew more slowly YoY at 11.0% compared to 15.9%
backsC15C16C17C18 - S09primarycompany filing2026-02-05Alphabet Form 10-K, fiscal year ended 31 December 2025 Alphabet Inc.the largest rival's own separation of performance from brand advertising, the products a mid-market merchant weighs against AppLovin Ads Manager, and the share of Alphabet that advertising is
Performance advertising lets our advertisers connect with users while driving measurable results.
backsC04 - S10primarycompany filing2026-01-29Meta Platforms Form 10-K, fiscal year ended 31 December 2025 Meta Platforms, Inc.the perimeter test for the largest cohort member and its own account of a self-service auction sold to performance advertisers
The majority of our marketers use our self-service ad platform to launch and manage their advertising campaigns.
read, but no claim rests on it - S11primaryvendor data2026-09-23Research pack for APP, frozen beside this thesis (refreshed 2026-09-23)content/APP/research-pack.jsonprice, the 252-day range and returns, the analyst split and its dated series, price-target drift, short interest, insider filings, the six restatements the filer made, the cohort multiples struck on the filed competitor set, the company's own six-year multiple history, and the split of the year's price move into revenue per share and multiplebacksC09C21C28C29C30C35C36C37C38C39C40C41C67
- S12primarycompany release2026-05-06the guidance range the June quarter was actually measured against, the restated first-quarter 2025 comparative, and the $1.0 billion first-quarter buyback
Second Quarter 2026 Financial Guidance Summary Revenue $1,915 $1,945 Adjusted EBITDA 1,615 1,645
backsC50C54 - S13primarycompany filing2026-08-05that the motion was still undecided at the last filing, and that a second set of proceedings is stayed behind it - so one ruling gates both. It also gives the negative: the legal proceedings note discloses no SEC inquiry, open or closed, anywhere.
The defendants filed a motion to dismiss the Amended Complaint in November 2025, and that motion was fully briefed as of February 2026. ... The Shareholder Derivative Complaints have been consolidated and stayed pending resolution of the defendants' motion to dismiss in the Brownback Action.
backsC51C55 - S14secondarynews2026-09-17the class period, the lead plaintiff deadline and the four alleged misstatements of the second securities class action, which no company filing has yet described
Rosen Law Firm ... announces a class action lawsuit on behalf of purchasers of securities of AppLovin Corporation (NASDAQ: APP) between February 12, 2026 and August 5, 2026, inclusive (the "Class Period"). ... If you wish to serve as lead plaintiff, you must move the Court no later than November 16, 2026. ... According to the lawsuit, throughout the Class Period, defendants made false and/or misleading statements and/or failed to disclose that: (1) the generative AI video creative feature for the company's AppLovin Ads platform was subject to significant development delays, making its release on AppLovin's timeline unlikely; (2) Defendants overstated the constancy with which AppLovin was improving its AI models; (3) for these reasons, among others, AppLovin had significantly overstated the benefits and reliability of the purportedly "virtuous cycle" and "compounding" value proposition that its AI models provided to customers and to AppLovin
backsC42C52 - S15primarycompany filing2026-04-07that the CTO since 2020 stepped down on 1 July 2026 to a non-executive Distinguished Engineer role, and that his successor, the Chief Product and Engineering Officer since November 2025, led the development of Axon 2.0
Basil Shikin informed the Company that he will step down from his role as Chief Technology Officer, effective as of July 1, 2026.
backsC53 - S16secondarynews2026-07-22the objections stated at the time of the early-2026 fall - Meta competing for untracked iOS traffic, Google's Project Genie for AI game creation, and a short-seller report from CapitalWatch - each reproduced verbatim on a second, independently phrased fetch
Meta Platforms was planning on competing for untracked ad traffic on Apple's iOS that it had historically ignored.
backsC54 - S17primaryvendor data2026-09-22the closing prices that date the three earlier drawdowns: 11 November 2021 to 27 December 2022, 14 February to 4 April 2025, and 22 December 2025 to 12 February 2026, and what followed eachbacksC54C55
- S18primaryindustry data2025-12-03AppsFlyer Performance Index, 2025 edition, press release AppsFlyer Ltd.an ordinal ranking of 88 mobile media sources from installs that an independent attribution firm measured for its own clients, 16.2 billion installs from 39,000 apps; the AppLovin sentence reproduced verbatim on a second fetch
Apple Ads remained the number one global media source for iOS gaming, with AppLovin narrowing the gap and leading in tier one North America and Western Europe.
backsC56 - S19primaryindustry data2024-10-14AppsFlyer Performance Index, 17th edition, press release AppsFlyer Ltd.the earlier reading from the same index, for the first half of 2024: second and third in gaming, from 13.5 billion non-organic installs of 35,000 apps
accelerated its ascent in the gaming realm, clinching second and third spots in the power and volume rankings
backsC56 - S20primarycompany filing2026-02-26Playtika Holding Corp. Form 10-K, fiscal year ended 31 December 2025 Playtika Holding Corp.a buyer's own account of what user acquisition costs it: $949.8 million of sales and marketing on $2,755.4 million of 2025 revenue, above its $758.5 million cost of revenue, bought from mobile ad networks, search and social on a payback rule
Our disciplined user acquisition strategy is centered on a payback period approach, which focuses on user monetization efforts that recoups our marketing spend during a reasonable timeframe.
backsC57 - S21primarycompany filing2026-04-21the beneficial ownership table at 31 March 2026 - the voting agreement at 66.9% of total voting power and the chief executive alone at 61.6% - and the executive officers as of that date
Shares subject to the Voting Agreement | 2,866,505 | * | 30,537,722 | 100 | 66.9
backsC41C45C53C61C62C63 - S22primarycompany filing2026-08-06Unity Software Form 10-Q, quarter ended 30 June 2026 Unity Software Inc.Grow Solutions revenue of $388,850 thousand in the June 2026 quarter against $287,162 thousand, and $740,441 thousand in the half against $571,784 thousand - so the first quarter too - with the growth credited to Unity Vector
primarily due to an increase in Grow Solutions revenue from growth in the Unity Ads Network, driven by “Unity Vector”, partially offset by decreases in revenue from the ironSource Ads Network
backsC58 - S23primarycompany filing2025-11-05Unity Software Form 10-Q, quarter ended 30 September 2025 Unity Software Inc.Grow Solutions for the third quarter and nine months of 2025 and 2024, from which the fourth quarter of each year follows by subtraction from the annual figures
Grow Solutions | 318,256 | 299,148 | 890,040 | 894,340
backsC58 - S24primarycompany filing2026-08-05Digital Turbine Form 10-Q, quarter ended 30 June 2026 Digital Turbine, Inc.the latest quarter of the second listed rival's demand-side business: $56,596 thousand, up 55.9%, mostly in Asia Pacific and China
App Growth Platform revenue increased by $20,304, or 55.9%, for the three months ended June 30, 2026 compared to the three months ended June 30, 2025, and was primarily driven by improved performance in the Asia Pacific and China regions
backsC14 - S25primarycompany filing2026-09-23every open-market sale with its plan flag - 579 sales of 1,064,402 shares, 52.4% of the shares under a plan, no purchases - the chief executive's discretionary sales in November, March and June, and the former CTO's 21 August notice of proposed sale
90,682 shares are being sold from Vasily SHIKIN individual
backsC41 - S26primarytranscript2026-05-06management's only sizing of the consumer business: the run rate a year earlier, and the spend it projects per new advertiser
So we talked about the $1 billion run rate a year ago. It's much bigger now. [...] Right now, we're projecting well over $70,000 a year from every new customer. So if you just want to size that, if we open up the platform and sign on 100,000 customers in the next year, first year revenue from them are ad spend -- advertising spend would be roughly $7 billion.
backsC65C66 - S27primarytranscript2026-02-11management declining to report the consumer business apart from gaming
On the breaking out e-commerce, we're not going to do that because we think of our platform as a unified platform.
backsC65 - S28primarytranscript2025-11-05the only figure for advertiser spend through the platform, against which revenue can be compared
what we reported, I think it was in Q1 was $11 billion plus of ad spend.
backsC68